Smoothness Prior Approach to Capturing Rapid Changes in Time-varying TFP and Application to the Chinese Economy

Abstract

In this paper we propose a Bayesian approach to statistical analysis of the economy with time-varying total factor productivity (TFP) showing rapid changes. Conventional approaches to the empirical study of economic growth lack flexibility and therefore cannot appropriately capture the complex movement of TFP. To solve this problem, we construct Bayesian models for a production function with dynamic structure. Also, the possibility is considered that there are rapid changes in TFP in some situations. In our proposed approach, TFP is treated as a time-varying parameter and regarded as a random variable. A set of Bayesian models for time-varying TFP incorporating rapid changes is constructed based on a smoothness prior approach. Then, the time-varying TFP is estimated using a Bayesian linear modeling approach together with a newly-proposed random grouping method. Using time series data for the Chinese macro-economy, we show that our proposed approach makes a detailed analysis of TFP trends possible.